Testing Method & Test Setup

You wouldn't want to use a pair of 8GB modules unless the rest of your platform is capable of putting it to good use. Luckily, we have got just a rig for the job.

Motherboard	Gigabyte Z77X-UD3H (BIOS F9)
CPU	Intel Core i7-3770K (ES, E0)
Graphic card	XFX 8600 GT
Memory	G.Skill TridentX F3-2400C10D-16GTX
HDD	Samsung 40 GB
PSU	Silverstone OP1000
OS	Windows 7, 64 bit SP1

From all the stability tests out there, we give our preference to HCI Memtest as, apart from time-consuming Prime 95 Blend, it's the toughest thing memory can face. To speed things up we will simultaneously use four instances, 3000Mb assigned each, and call things stable if we see four progress bars going past 100% error-free.

Results

First, we discovered that ability of this memory to run certain tRCD and tRP values at certain frequency is not dependent on other timings or voltage - the limitations for each are more or less constant, as displayed by the right part of the diagram.
The other thing we've noticed is that, provided tRCD and tRP are not the limiting factors, voltage scaling using different CAS values is almost linear. This phenomena is typical for most of modern ICs and it is quite helpful in a way that knowing stable operating voltage for certain settings allows you to give an estimate to what sort of voltage you'll need for a, say, 100MHz frequency bump without having to guess.
Having explained how things work, we can see that there isn't much overclocking headroom on these Trident-X modules. As all our attempts to make them work above 1230MHz resulted in a failure and once even in a corrupt OS, the best thing a user can do is lowering the timings to 10-10-12 and the voltage to 1.60V at the rated frequency. While such overclockability is still impressive for 4Gb density ICs, people who will get such on single-sided 4GB modules have their full rights to complain.

Page 1 - Introduction	Page 4 - Results
Page 2 - Closer Look	Page 5 - Conclusion
Page 3 - Photo Gallery

Discuss this article in the forums